Intelligent slice image segmentation processing method based on clinical problems

Abstract

The invention relates to the technical field of image processing, in particular to a slice image intelligent segmentation processing method based on clinical problems. The method comprises: acquiring a slice image of a medical image; a window width is set, an optimal segmentation threshold value under the window width is obtained through an optimal segmentation threshold value obtaining step, image segmentation is carried out on the slice image according to the optimal segmentation threshold value, and an optimal segmentation image is obtained; changing the window width, and obtaining corresponding optimal segmentation images under different window widths by using an optimal segmentation threshold value obtaining step; and for the preferable segmented image under each window width, calculating a segmentation effect value corresponding to the preferable segmented image, and screening out the optimal segmented image from the preferable segmented images. According to the method, the optimal segmented image is screened based on the calculated segmentation effect values corresponding to the optimal segmented image under different window widths. And obtaining an optimal segmentation image on the basis of the segmentation effect value of the optimal segmentation image, thereby realizing the optimal segmentation of the slice image.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to an intelligent segmentation processing method for slice images based on clinical problems. Background Art

[0002] Image processing plays an important role in medical diagnosis and clinical treatment, etc. Generally, it is to process medical images such as CT images and magnetic resonance imaging images collected. The processing of medical images is usually to be able to more clearly analyze the lesions in the medical images. Therefore, the segmentation of the lesion area in the medical images is crucial.

[0003] Currently, a common method for segmenting the lesion area in medical images is to use a fuzzy threshold segmentation algorithm to segment the lesion area. Among them, the selection of the window width of the membership function in the fuzzy threshold segmentation algorithm is: select the window width of the membership function corresponding to the medical image according to the distance between the peaks of the histogram corresponding to the medical image. When selecting the window width of the membership function by this method, it is impossible to judge the segmentation effect when using the selected window width to segment the lesion area in the medical image. When the segmentation effect is poor, it will be difficult to clearly segment the lesion area. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide an intelligent segmentation processing method for slice images based on clinical problems. The specific technical solutions adopted are as follows: Obtain the slice image of the medical image; Set the window width, use the preferred segmentation threshold obtaining step to obtain the preferred segmentation threshold at this window width, and perform image segmentation on the slice image with this preferred segmentation threshold to obtain a preferred segmentation image; change the window width, and use the preferred segmentation threshold obtaining step to obtain the preferred segmentation thresholds and the corresponding preferred segmentation images at different window widths; For the preferred segmentation image at each window width, obtain the segmentation obviousness of the target area according to the pixel value difference between each edge pixel point of the target area segmented in the preferred segmentation image and its corresponding neighborhood pixel point; obtain the image structure similarity of the preferred segmentation image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target area; combine the segmentation obviousness and the image structure similarity to obtain a segmentation effect value, and determine the best segmentation image from the preferred segmentation images according to the segmentation effect value.

[0005] Preferably, the obtaining of the segmentation obviousness of the target area according to the pixel value difference between each edge pixel point of the target area segmented in the preferred segmentation image and its corresponding neighborhood pixel point includes: The calculation formula of the segmentation obviousness is: Among them, is the obviousness of the segmentation; is the natural constant; is the number of target regions in the segmented image; is the standard deviation of the pixel values of the edge pixels of all target regions; is the maximum value of the pixel value differences between all edge pixels in the i-th target region in the segmented image and the pixels in their corresponding eight-neighborhoods; is the maximum value of the pixel value differences between the edge pixels in all target regions in the segmented image and the pixels in their corresponding eight-neighborhoods; is the minimum value of the pixel value differences between the edge pixels in all target regions in the segmented image and the pixels in their corresponding eight-neighborhoods; is the maximum difference of the adjacent pixel value differences.

[0006] Preferably, the method for obtaining the maximum difference of the adjacent pixel value differences is: obtaining the pixel value differences between the edge pixels in all target regions in the segmented image and the pixels in their corresponding eight-neighborhoods, sorting the pixel value differences in ascending order to obtain a pixel value difference sequence; the maximum difference between the adjacent pixel value differences in the pixel value difference sequence is the maximum difference of the adjacent pixel value differences.

[0007] Preferably, the steps for obtaining the preferred segmentation threshold include: Performing threshold segmentation on the slice image based on different fuzzy segmentation thresholds to obtain multiple segmented images; obtaining the membership degree of each pixel point in each segmented image based on the membership function with a set window width, calculating the fuzzy rate of the segmented image from the membership degree; establishing a segmentation threshold - fuzzy rate curve, and the fuzzy segmentation threshold corresponding to the bottom of the segmentation threshold - fuzzy rate curve is the preferred segmentation threshold.

[0008] Preferably, the image structure similarity of the preferred segmented image obtained from the mean of the structural similarities of each sliding window when the sliding window slides across the target region includes: For any target region, calculating the structural similarity of each sliding window when the sliding window slides across the target region, and the mean of the structural similarities of each sliding window is the regional structure similarity of the target region; Calculating the mean of the regional structure similarities corresponding to all target regions as the initial image structure similarity; using the exponential function with the natural constant as the base and the minimum regional structure similarity as the exponent as the similarity adjustment parameter; determining the image structure similarity according to the initial image structure similarity and the similarity adjustment parameter.

[0009] Preferably, determining the image structure similarity according to the initial image structure similarity and the similarity adjustment parameter includes: The product of the initial image structure similarity and the similarity adjustment parameter is the image structure similarity.

[0010] Preferably, obtaining the segmentation effect value by combining the segmentation distinctness and the image structure similarity includes: The sum of the segmentation distinctness and the image structure similarity is the initial segmentation effect value; the exponential function with the natural constant as the base and the initial segmentation effect value as the exponent is the segmentation effect value.

[0011] Preferably, the set window width is: the window width of the set membership function.

[0012] Preferably, establishing the segmentation threshold - fuzziness rate curve includes: According to the fuzziness rate of the segmented image corresponding to different fuzzy segmentation thresholds, establish a segmentation threshold - fuzziness rate curve with the fuzzy segmentation threshold as the abscissa and the fuzziness rate as the ordinate.

[0013] Preferably, determining the optimal segmentation image from the preferred segmentation images according to the segmentation effect value includes: Taking the preferred segmentation image with the largest segmentation effect value as the optimal segmentation image.

[0014] The embodiments of the present invention have at least the following beneficial effects: This method obtains the slice image of the medical image; sets the window width, uses the preferred segmentation threshold obtaining step to obtain the preferred segmentation threshold at this window width, and performs image segmentation on the slice image with this preferred segmentation threshold to obtain the preferred segmentation image; changes the window width, uses the preferred segmentation threshold obtaining step to obtain the preferred segmentation thresholds and the corresponding preferred segmentation images at different window widths; uses the fuzzy threshold segmentation algorithm, by changing the fuzzy segmentation threshold in the fuzzy operation parameters, obtains the segmentation effect of the image, and screens out the preferred segmentation image with the best segmentation effect at the same preferred width.

[0015] For the preferred segmented images at each window width, calculate the segmentation effect value of each preferred segmented image according to the segmentation clarity of the edges of the target region in the preferred segmented image, and determine the best segmented image from the preferred segmented images according to the segmentation effect value. Since there may be some differences between the pixel values of the target region and the background region, the target region obtained by the fuzzy threshold segmentation algorithm may have the problem of unclear segmentation. Therefore, calculate the segmentation effect value of the preferred segmented images after fuzzy threshold segmentation at different window widths, and use the segmentation effect value as the standard for judging the segmentation quality of the preferred segmented images to obtain the best segmented image with the best segmentation effect. The present invention uses the fuzzy threshold segmentation algorithm, obtains the segmentation effect of the image by changing the fuzzy operation parameters, and finally determines the best segmentation effect to realize the intelligent segmentation of clinical section images. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a flowchart of a method for intelligent segmentation and processing of section images based on clinical problems provided by an embodiment of the present invention. Detailed Embodiments

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following will, in conjunction with the accompanying drawings and preferred embodiments, describe in detail the specific embodiments, structures, features and effects of a method for intelligent segmentation and processing of section images based on clinical problems proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0020] An embodiment of the present invention provides a specific implementation method of an intelligent segmentation processing method for slice images based on clinical problems, which is applicable to the segmentation scenario of slice images in the medical scenario. In this scenario, the slice images corresponding to the medical images captured by professional medical devices are segmented. In order to solve the problem that the window width of the membership function corresponding to the medical image is selected according to the distance between the peaks of the histogram corresponding to the medical image, and it is impossible to judge the segmentation effect when using the selected window width to segment the lesion area of the medical image, which will lead to difficulty in clearly segmenting the lesion area when the segmentation effect is poor. In the embodiment of the present invention, after using the fuzzy threshold segmentation algorithm to change the fuzzy segmentation threshold and select the optimal segmentation image with the best segmentation effect under the same optimal width, based on the calculated segmentation effect values corresponding to the optimal segmentation images under different window widths, the best segmentation image is selected. Based on the segmentation effect value of the optimal segmentation image, the best segmentation image is obtained to achieve the best segmentation of the slice image.

[0021] The following specifically describes the specific solution of an intelligent segmentation processing method for slice images based on clinical problems provided by the present invention with reference to the accompanying drawings.

[0022] Please refer to Figure 1 , which shows a flowchart of the steps of an intelligent segmentation processing method for slice images based on clinical problems provided by an embodiment of the present invention. The method includes the following steps: Step S100, obtain the slice image of the medical image.

[0023] The medical image is sliced to obtain the slice image corresponding to the medical image. Among them, the medical image is an image containing a lesion area captured by medical staff through a professional medical device. The obtained slice image is grayscale processed, and the slice image in the subsequent steps is the grayscale processed slice image.

[0024] Step S200, set the window width, use the optimal segmentation threshold obtained in the previous step to obtain the optimal segmentation threshold under this window width, and segment the slice image with this optimal segmentation threshold to obtain the optimal segmentation image; change the window width, and use the optimal segmentation threshold obtained in the previous step to obtain the optimal segmentation thresholds and corresponding optimal segmentation images under different window widths.

[0025] The main purpose of segmenting the slice image corresponding to the medical image is to target the lesion area in the slice image and segment out the lesion area. Generally, the more obvious the characteristics of the lesion area are, the easier it is to be segmented out. However, due to the diverse causes of lesions in actual situations and the diverse forms of lesion areas, it is difficult to accurately segment out the lesion area only through threshold segmentation. The present invention uses fuzzy threshold segmentation to segment the lesion area, and obtains the best segmentation effect by changing the window width of the membership function and the fuzzy segmentation threshold, and obtains the best segmentation image corresponding to the best segmentation effect, realizing intelligent image segmentation.

[0026] Since the information in the medical image is relatively complex, when the characteristics of the lesion area are not very obvious, the contrast between the lesion area and the normal area is not strong, and there is an overlapping part between the pixel values of the pixel points in the lesion area and the pixel values of the pixel points in the normal area, it is considered that there is a certain similarity and uncertainty between the pixel values of the lesion area and the normal area at this time. Therefore, direct use of threshold segmentation cannot well segment out the lesion area. In image segmentation, when the contrast between the segmented target area and the background area is not obvious, and the pixel values of the target area and the background area have similarity and uncertainty, fuzzy theory is usually used for representation and segmentation is performed according to the fuzzy threshold segmentation algorithm.

[0027] When using the fuzzy threshold segmentation algorithm to segment the slice image of the medical image to be segmented, the window width of the membership function and the fuzzy segmentation threshold in the fuzzy threshold segmentation algorithm are respectively selected so that the obtained segmented image is the best segmentation image that can clearly segment out the lesion area.

[0028] First, set the window width w of the membership function, obtain the optimal segmentation threshold under this window width w by using the optimal segmentation threshold obtaining step, and perform image segmentation on the slice image with this optimal segmentation threshold to obtain the optimal segmentation image. It should be noted that the optimal segmentation image corresponding to each set window width is the clearest segmentation image that segments out the lesion area under this window width, that is, one optimal segmentation image corresponding to each set window width.

[0029] Among them, the optimal segmentation threshold obtaining step is as follows: Step 1, based on different fuzzy segmentation thresholds, perform threshold segmentation on the slice image to obtain multiple segmented images. Based on the membership function with the set window width, obtain the membership degree of each pixel point in each segmented image, and calculate the fuzzy rate of the segmented image from the membership degree.

[0030] Set the size of the slice image to , and the number of gray levels in the slice image is L.

[0031] The membership degree of the pixel point with the gray value of in the segmented image is expressed as , where . An expression for constructing the fuzziness rate of the segmented image is needed. It should be noted that the pixel value in the grayscale image is the grayscale value.

[0032] The calculation formula for the constructed fuzziness rate is: where is the fuzziness rate; is the number of pixel points with grayscale value in the segmented image; is the length of the segmented image; is the width of the segmented image; is the membership degree of the pixel point with grayscale value ; is the number of grayscale levels of the sliced image; is the minimum value function; is and the smaller value of. It should be noted that the number of grayscale levels corresponding to the sliced image and the segmented image is the same. It should be noted that the calculation of this fuzziness rate is prior art and will not be elaborated here.

[0033] In the calculation of the fuzzification degree of the image, different membership functions will result in different fuzzification degrees. Therefore, in the calculation of the fuzzification degree of different segmented images, the corresponding membership function needs to be set according to the characteristics of the segmented image. In the present invention, the membership function selects the existing standard S-type function.

[0034] In the membership function, the window width is set to . From the calculation formula of the fuzziness rate, it can be seen that at this time, the size of the fuzziness rate is only related to the size of the membership function. According to the calculation formula of the membership function, the membership function is determined by the window width and the parameter q. Once the window width is determined in the membership function, the fuzziness rate is only related to the parameter q. Thus, the fuzziness rate curve can be affected by traversing the parameter q in the grayscale interval, thereby determining the selection of the fuzzy segmentation threshold. Therefore, the selection of the window width and the parameter q is a decisive factor for the image segmentation effect, and this parameter q is the fuzzy segmentation threshold.

[0035] In the process of selecting the fuzzy segmentation threshold, first set the window width w. Under the condition that the window width is temporarily determined, based on the membership function with the set window width, obtain the membership degree of each pixel point in each segmented image, and calculate the fuzziness rate of the segmented image from the membership degree.

[0036] Step 2: Establish a segmentation threshold - fuzziness ratio curve. The fuzzy segmentation threshold corresponding to the bottom of the segmentation threshold - fuzziness ratio curve is the optimal segmentation threshold.

[0037] Change the fuzzy segmentation threshold so that the membership function slides within the range of [1, L] according to the change of the fuzzy segmentation threshold, and calculate the fuzziness ratio of the segmented image corresponding to each different fuzzy segmentation threshold. According to the fuzziness ratios of the segmented images corresponding to different fuzzy segmentation thresholds, establish a segmentation threshold - fuzziness ratio curve with the fuzzy segmentation threshold as the abscissa and the fuzziness ratio as the ordinate. Find the bottom of this segmentation threshold - fuzziness ratio curve, and use the fuzzy segmentation threshold corresponding to this bottom as the optimal segmentation threshold when the window width is w.

[0038] Further, change the window width, and use the optimal segmentation threshold obtained in the previous step to obtain the optimal segmentation thresholds and the corresponding optimal segmented images under different window widths.

[0039] Analyze the calculation process of the fuzzy segmentation threshold. The membership function directly affects the segmentation threshold - fuzziness ratio curve, and thus affects the fuzzy segmentation threshold. In the present invention, for a standard membership function with a determined form, the window width directly affects the change of the membership function. Therefore, the window width plays a key role in determining the fuzzy segmentation threshold. By analyzing the formation of the segmentation threshold - fuzziness ratio curve, judge the influence of the value of the window width on the final fuzzy segmentation threshold: the smaller the value of the window width , the steeper the membership function curve, which will cause the fuzziness ratio to oscillate near the bottom, thus generating false thresholds; the larger the window width , the flatter the membership function curve, which may smooth out the bottom of the segmentation threshold - fuzziness ratio curve, ultimately resulting in the loss of the candidate fuzzy segmentation threshold. Therefore, the selection of the window width of the membership function in the fuzzy threshold segmentation algorithm is crucial.

[0040] Step S300: For the optimal segmented image under each window width, obtain the segmentation obviousness of the target region according to the pixel value differences between each edge pixel point of the target region segmented from the optimal segmented image and its corresponding neighborhood pixel points; obtain the image structure similarity of the optimal segmented image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target region; combine the segmentation obviousness and the image structure similarity to obtain a segmentation effect value, and determine the best segmented image from the optimal segmented images according to the segmentation effect value.

[0041] By setting different window widths, different optimal segmentation thresholds can be obtained. For different lesion regions, the most suitable window widths are different. Therefore, the most appropriate window width is selected according to the image features. In the present invention, different window widths are selected , and by analyzing the segmentation effects corresponding to different window widths, the best segmentation effect is obtained. The steps for obtaining the best segmentation image with the best segmentation effect are specifically as follows: Step 1: For the optimal segmentation images under each window width, the segmentation obviousness of the target region is obtained according to the pixel value differences between the edge pixel points of the target region segmented from the optimal segmentation image and their corresponding neighborhood pixel points.

[0042] First, obtain the optimal segmentation images corresponding to each different window width and the target regions in the optimal segmentation images. The target region is the segmented lesion region. Since the edge of the segmented target region has a higher contrast relative to the background region. At the same time, there may be an intersection between the pixel values of the target region and the background region. Therefore, there may be misjudgment in the target region obtained by the fuzzy threshold segmentation algorithm, that is, the pixel points in the background region that intersect with the pixel values of the target region are segmented as the target region. Therefore, at this time, according to the characteristics of the segmented target region, the current segmentation obviousness effect is judged. That is, for the optimal segmentation images under each window width, the segmentation obviousness of the target region is obtained according to the pixel value differences between the edge pixel points of the target region segmented from the optimal segmentation image and their corresponding neighborhood pixel points.

[0043] The formula for the segmentation obviousness is as follows: where is the natural constant; is the number of target regions in the segmentation image; is the standard deviation of the pixel values of the edge pixel points of all target regions; is the maximum value of the pixel value differences between all edge pixel points in the i-th target region in the segmentation image and their corresponding eight-neighborhood pixel points; is the maximum value of the pixel value differences between the edge pixel points of all target regions in the segmentation image and their corresponding eight-neighborhood pixel points; is the minimum value of the pixel value differences between the edge pixel points of all target regions in the segmentation image and their corresponding eight-neighborhood pixel points; is the maximum difference between adjacent pixel value differences.

[0044] Among them, the method for obtaining the maximum difference of adjacent pixel value differences is as follows: obtain the pixel value differences between the edge pixel points in all target regions in the segmented image and the pixel points in their corresponding eight-neighborhoods, and sort the pixel value differences in ascending order to obtain a pixel value difference sequence; the maximum difference between adjacent pixel value differences in the pixel value difference sequence is the maximum difference of adjacent pixel value differences. It should be noted that the edge pixel points are the pixel points on the edge of the target region.

[0045] In the calculation formula of the segmentation distinctness degree, is the mean value of the pixel value differences between the edge pixel points in all target regions in the segmented image and the pixel points in their corresponding eight-neighborhoods. Generally, the edges of the target regions in the segmented image have a large contrast, so the pixel value differences between the corresponding edge pixel points and the pixel points in their corresponding eight-neighborhoods are large. Therefore, the larger the value, the better the segmentation effect of the current segmented image, the more distinct the segmentation of the lesion region, and the greater the corresponding segmentation distinctness degree. The standard deviation of the pixel values of the edge pixel points of all target regions represents the pixel difference of the edge pixel points of the target region. For the same type of target region, the pixel value differences between its edge pixel points and the pixel points in its neighborhood are small, then the corresponding is smaller, which reflects the better current segmentation distinctness effect, that is, the larger the value of the segmentation distinctness degree. For accurate target regions, the pixel value differences of the edge pixel points of each target region are similar. Therefore, when there is a misjudgment of the target region, the pixel value differences of the edge pixel points of the misjudged region will deviate from the pixel value differences of the edge pixel points of the normal target region. represents the range value of the pixel value differences of the edge pixel points in all target regions. The smaller this range value, the better the segmentation distinctness effect. The maximum difference of adjacent pixel value differences reflects the deviation degree of the pixel value differences of the edge pixel points of the target region. The larger the maximum difference of its pixel value differences, the greater the corresponding deviation degree, which reflects the poorer current segmentation effect and the smaller the corresponding segmentation distinctness degree.

[0046] Step 2: Obtain the image structure similarity of the optimal segmented image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target region.

[0047] For any target region, calculate the structural similarity of each sliding window when the sliding window slides across the target region. The mean value of the structural similarities of each sliding window is the regional structural similarity of the target region.

[0048] For the target segmentation in the slice image, the accurate target regions required are all the same type of lesion regions, so the target regions have a high similarity. First, calculate the sizes of all segmented target regions. The size of the smallest target region is , with the size of the smallest target area as the size of the sliding window.

[0049] Perform a sliding window operation on all the segmented target areas. Calculate the structural similarity of the area corresponding to each sliding window in each target area. Since the similarity between accurate target areas is high, the structural similarity of the corresponding sliding windows should also be relatively high. For the th target area in the segmented image, calculate the structural similarity of each sliding window when the sliding window passes through this target area. The average value of the structural similarities of each sliding window is the regional structural similarity of this target area.

[0050] The th target area's regional structural similarity The calculation formula is: Wherein, is the number of areas corresponding to the sliding window in the th target area, is the structural similarity of the area corresponding to the th sliding of the sliding window in the th target area.

[0051] This regional structural similarity is the average value of the structural similarities of the areas corresponding to the sliding windows within the target area. The greater the structural similarity of the areas corresponding to the sliding windows within the target area, the greater the regional structural similarity of the corresponding target area. It should be noted that the calculation of the structural similarity is prior art and will not be elaborated here.

[0052] Then, for all target regions in the segmented image, calculate the regional structural similarity of all target regions obtained by the current segmentation. Calculate the mean of the regional structural similarities corresponding to all target regions as the initial image structural similarity. The larger the initial image structural similarity, the higher the similarity between the target regions obtained by the current segmented image. Further, to highlight the similarity of the segmented regions, select the target region with the smallest regional structural similarity to represent the least similar region, reflecting the similarity of the segmented target regions. The larger the smallest regional structural similarity, the greater the similarity of the target regions in the current segmented image. At the same time, taking the natural constant as the base and the smallest regional structural similarity as the exponent, the exponential function is used as the similarity adjustment parameter. The smallest regional structural similarity represents the least similar target region among all target regions, directly reflecting the regional similarity of the target regions in the entire segmented image with the worst similarity. Then, use the exponential function to amplify the smallest regional structural similarity, so that the similarity adjustment parameter can more directly adjust the overall image structural similarity of the segmented image. Adjust the initial image structural similarity according to the similarity adjustment parameter, that is, the product of the initial image structural similarity and the similarity adjustment parameter is the final image structural similarity. The image structural similarity adjusted by the similarity adjustment parameter can more intuitively reflect the overall image structural similarity of the segmented image. The larger this image structural similarity, the greater the similarity degree of each target region in the segmented image, and the more accurate the segmentation of the target region by the segmented image, that is, the more accurate the segmentation of the lesion region by the segmented image.

[0053] Step 3: Combine the segmentation obviousness and the image structural similarity to obtain the segmentation effect value. The preferred segmented image with the largest segmentation effect value is the best segmented image.

[0054] Obtain different window widths in the fuzzy threshold segmentation operation The segmentation obviousness and the image structural similarity corresponding to the preferred segmentation threshold. The segmentation obviousness and the image structural similarity reflect the segmentation obvious effect and the segmentation accuracy of the preferred segmentation threshold. Combine the segmentation obviousness and the image structural similarity to obtain the segmentation effect value that can reflect the final fuzzy threshold segmentation effect. Specifically: add the segmentation obviousness and the image structural similarity to obtain the initial segmentation effect value; taking the natural constant as the base and the initial segmentation effect value as the exponent, the exponential function is the segmentation effect value.

[0055] This segmentation effect value The calculation formula of is: Among them, is the segmentation obviousness; is the image structural similarity; is the initial segmentation effect value; is the natural constant.

[0056] Among them, both the segmentation distinctness and the image structure similarity are directly proportional to the segmentation effect value. The greater the segmentation distinctness, the greater the corresponding segmentation effect value; the greater the image structure similarity, the greater the corresponding segmentation effect value. The greater the segmentation effect value, the better the current segmentation effect.

[0057] In the fuzzy threshold segmentation operation of the slice image, the membership function in the slice image is mainly constructed. In the present invention, a standard S-shaped function is selected as the membership function. At this time, in the process of obtaining the optimal fuzzy segmentation threshold, the selected window width directly affects the final fuzzy segmentation threshold. Construct the expression of the segmentation effect value. Different window widths obtain different segmentation effect values. Each preferred segmentation image under each window width corresponds to a segmentation effect value. The segmentation effect value The greater it is, the better the segmentation effect. Select the preferred segmentation image with the largest segmentation effect value as the optimal segmentation image. The window width corresponding to this optimal segmentation image is the optimal window width, and the corresponding preferred segmentation threshold is the optimal segmentation threshold. Therefore, in the actual fuzzy threshold segmentation algorithm, by changing the value of the window width, different segmentation effect values are obtained. On the basis of changing the window width, the maximum value of the segmentation effect value F is obtained by using the existing technology. For example, on the basis of changing the window width, the simulated annealing algorithm is used to obtain the maximum segmentation effect value. At this time, the corresponding segmentation effect is the best.

[0058] In medical images, the segmentation of the lesion area is beneficial to the analysis of the cause and degree of the lesion. The fuzzy threshold segmentation algorithm is used to segment the lesion area. By changing the size of the window width, the optimal segmentation image corresponding to the optimal segmentation effect is obtained, that is, the target area segmented in this optimal segmentation image is the most real and accurate lesion area, realizing the intelligent segmentation of the slice image in clinical medicine.

[0059] In summary, the present invention relates to the technical field of image processing. The method obtains a slice image of a medical image; sets a window width, obtains an optimal segmentation threshold at the window width by using an optimal segmentation threshold obtaining step, and performs image segmentation on the slice image with the optimal segmentation threshold to obtain an optimal segmentation image; changes the window width, and obtains the optimal segmentation thresholds and corresponding optimal segmentation images at different window widths by using the optimal segmentation threshold obtaining step; for the optimal segmentation image at each window width, obtains the segmentation obviousness of the target region according to the pixel value difference between each edge pixel point of the target region segmented in the optimal segmentation image and its corresponding neighborhood pixel point; obtains the image structure similarity of the optimal segmentation image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target region; combines the segmentation obviousness and the image structure similarity to obtain a segmentation effect value, and the optimal segmentation image with the largest segmentation effect value is the best segmentation image. The present invention uses a fuzzy threshold segmentation algorithm, obtains the segmentation effect of the image by changing the fuzzy operation parameters, and finally determines the best segmentation effect to realize the intelligent segmentation of clinical slice images.

[0060] It should be noted that the above sequence of embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0061] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0062] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent segmentation processing method for slice images based on clinical problems, characterized in that, The method includes the following steps: Obtain sliced images of medical images; Set a window width, use the preferred segmentation threshold obtained in the preferred segmentation threshold obtaining step to perform image segmentation on the sliced image at this window width to obtain a preferred segmentation image; change the window width, and use the preferred segmentation threshold obtained in the preferred segmentation threshold obtaining step to obtain the preferred segmentation thresholds and corresponding preferred segmentation images at different window widths; For the preferred segmentation image at each window width, obtain the segmentation distinctness of the target region according to the pixel value differences between the edge pixel points of the target region segmented from the preferred segmentation image and their corresponding neighborhood pixel points; obtain the image structure similarity of the preferred segmentation image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target region; combine the segmentation distinctness and the image structure similarity to obtain a segmentation effect value, and determine the best segmentation image from the preferred segmentation images according to the segmentation effect value.

2. The intelligent segmentation processing method for slice images based on clinical problems according to claim 1, wherein The obtaining of the segmentation distinctness of the target region according to the pixel value differences between the edge pixel points of the target region segmented from the preferred segmentation image and their corresponding neighborhood pixel points includes: The calculation formula of the segmentation distinctness is: wherein, is the obvious degree of the segmentation; is the natural constant; is the number of target regions in the segmented image; is the standard deviation of the pixel values of the edge pixels of all target regions; is the maximum value of the pixel value differences between all edge pixels in the i-th target region in the segmented image and the pixels in its corresponding eight-neighborhood; is the maximum value of the pixel value differences between the edge pixels in all target regions in the segmented image and the pixels in their corresponding eight-neighborhood; is the minimum value of the pixel value differences between the edge pixels in all target regions in the segmented image and the pixels in their corresponding eight-neighborhood; is the maximum difference between adjacent pixel value differences.

3. The intelligent segmentation processing method for slice images based on clinical problems according to claim 2, characterized in that The method for obtaining the maximum difference of adjacent pixel value differences is: obtain the pixel value differences between the edge pixel points in all target regions in the segmentation image and the pixel points in the corresponding eight-neighborhood, and sort the pixel value differences in ascending order to obtain a pixel value difference sequence; the maximum difference between adjacent pixel value differences in the pixel value difference sequence is the maximum difference of adjacent pixel value differences.

4. A method for intelligent segmentation and processing of slice images based on clinical problems according to claim 1, characterized in that, The preferred segmentation threshold obtaining step includes: Based on different fuzzy segmentation thresholds, perform threshold segmentation on the sliced image to obtain multiple segmentation images; based on the membership function of the set window width, obtain the membership degree of each pixel point in each segmentation image, and calculate the fuzzy rate of the segmentation image from the membership degree; establish a segmentation threshold - fuzzy rate curve, and the fuzzy segmentation threshold corresponding to the bottom of the segmentation threshold - fuzzy rate curve is the preferred segmentation threshold.

5. A method for intelligent segmentation processing of slice images based on clinical problems according to claim 1, characterized in that, The obtaining of the image structure similarity of the preferred segmentation image from the mean value of the structural similarities of each sliding window when the sliding window slides across the target region includes: For any target region, calculate the structural similarity of each sliding window when the sliding window slides across the target region, and the mean value of the structural similarities of each sliding window is the regional structure similarity of the target region; Calculate the mean value of the regional structure similarities corresponding to all target regions as the initial image structure similarity; use the exponential function with the natural constant as the base and the minimum regional structure similarity as the exponent as the similarity adjustment parameter; determine the image structure similarity according to the initial image structure similarity and the similarity adjustment parameter.

6. The intelligent segmentation processing method for slice images based on clinical problems according to claim 5, wherein, The determining of the image structure similarity according to the initial image structure similarity and the similarity adjustment parameter includes: The product of the initial image structure similarity and the similarity adjustment parameter is the image structure similarity.

7. A method for intelligent segmentation and processing of slice images based on clinical problems according to claim 1, characterized in that The combining of the segmentation distinctness and the image structure similarity to obtain a segmentation effect value includes: The sum of the segmentation distinctness degree and the image structure similarity gives the initial segmentation effect value; the exponential function with the natural constant as the base and the initial segmentation effect value as the exponent is the segmentation effect value.

8. A method for intelligent segmentation and processing of slice images based on clinical problems according to claim 1, characterized in that, The set window width is: the window width of the set membership function.

9. A method for intelligent segmentation and processing of slice images based on clinical problems according to claim 4, characterized in that, The establishment of the segmentation threshold - fuzziness rate curve includes: Based on the fuzziness rate of the segmented images corresponding to different fuzzy segmentation thresholds, a segmentation threshold - fuzziness rate curve with the abscissa being the fuzzy segmentation threshold and the ordinate being the fuzziness rate is established.

10. A method for intelligent segmentation and processing of slice images based on clinical problems according to claim 1, characterized in that, The determination of the optimal segmentation image from the preferred segmentation images according to the segmentation effect value includes: Taking the preferred segmentation image with the largest segmentation effect value as the optimal segmentation image.

Images